50's: The Algorithmic Composition and The Illiac Suite
The first attempts at computer-generated music appeared in the 1950s with a focus on algorithmic music creation. The advent of computer-generated music by pioneers such as Alan Turing with the Manchester Mark II computer opened up multiple possibilities for research in musical intelligence where computer systems could recognize, create, and analyze music.
Turing helped create one of the first computer music recordings by repurposing signal sounds on one of Manchester's first computers.
Photo from the Science & Society Image Library.
Specifically, early experiments focused on algorithmic composition (a computer that uses formal sets of rules to create music). In 1957, we saw the first work composed exclusively by artificial intelligence — Illiac Suite for String Quartet.
​
Through the use of mathematical models and algorithms, Lejaren Hiller (an American composer) and Leonard Isaacson (an American composer and mathematician) created Illiac Suite, the first original piece composed by a computer. To achieve this feat, they used a Monte Carlo algorithm that generated random numbers that corresponded to certain musical characteristics such as pitch or rhythm.
​
Using a set of constraints, these random features were limited to elements that would be musically 'cool' as defined by the rules of traditional music theory, statistical probabilities (such as Markov chains), and the imagination of the two composers.
Lejaren Hiller and Leonard Isaacson.
Photo by Illinois Distributed Museum
60's: Stochastic Probabilities and Pattern-Recognizing Computer
Another innovator in this field was Iannis Xenakis, a composer and engineer, who used stochastic probabilities to aid in the creation of his music. A stochastic process is a mechanism with random probability distributions that cannot be predicted but can be analyzed statistically. In the early 1960s, he used computers and the FORTRAN language to weave together multiple probability functions to determine the overall structure and other parameters (such as pitch and dynamics) of a composition.
Iannis Xenakis in Théâtre du Châtelet, Paris.
Xenakis modeled his music as if he were modeling a science experiment. Each instrument was like a molecule and would go through its own stochastic, random process to determine its behavior (the frequency of pitch and the velocity of certain notes). His work introduced new methods for creating sound, but it also served as one of the first examples of AI functioning as a complementary analysis tool rather than just a compositional tool. The way Xenakis created his melodies and orchestrations for different instruments was inspired by the sound spaces shaped by the stochastic process.
​
In 1965, inventor Ray Kurzweil debuted a piano piece created by a computer capable of recognizing patterns in various compositions. The computer was then able to analyze and use these patterns to create new melodies. The computer debuted on the quiz show I've Got a Secret and left hosts baffled until movie star Henry Morgan guessed Ray's secret.
Ray Kurzweil presenting his creation to a group of students.
From Musical Robots to Generative Intelligence
A musical robot bears a resemblance to the early experiments of the 1950s and 60s; it can recognize patterns, possesses a musical grammar, and exhibits a general problem-solving sense. However, it achieves its goals through rather direct and forceful methods. On the other hand, musical intelligence replaces the brute force approach of the robot with a knowledge-based understanding system, having its own awareness of how musical elements can function. This trend of AI systems constructing their own self-sufficient understanding of musical elements laid the foundation for the more sophisticated musical intelligence we observe today.
Further developments in this period continued to push the boundaries of computational creativity. For instance, Robert Rowe devised a system wherein a machine could infer the metric, tempo, and note durations as someone freely plays on a keyboard. In 1995, Imagination Engines trained a neural network with popular melodies, employing reinforcement learning, leading to the generation of over 10,000 new musical choruses. Reinforcement learning involves training a neural network to achieve a goal, rewarding/punishing the model based on the decisions it makes to reach a specified objective.
_edited.jpg)
In 2007, Imagination Engines released the album “Song of the Neurons”, a collection of 14 musical compositions written by a Creativity Machine, without resorting to the methods used by computational musicians (i.e., rules and templates. Instead, this neural architecture improved itself to impressively high levels of musical skill by simply observing the facial expressions of Thaler, the founder of the organization, as it generated candidate melodies.
